design and implementation of automatic system …failure scenario 5 selection of recipes from db...
TRANSCRIPT
Design and Implementation of Automatic System for Network Testing with Quality Degradation
Sep. 27, 2017, Seoul, Korea
Junichi Kawasaki,
Megumi Shibuya, Atsuo Tachibana, Masanori Miyazawa and Teruyuki Hasegawa
KDDI Research, Inc.
1. Introduction
Network Testing
SDN/NFV
×
Service Design Configuration Testing S-inLegacy
SDN/NFV Test items
Manual-based approach- cannot cover all items- human error risk
-To maintain SLA between customers and carriers
-When devices are added, removed, upgraded
-Over hundreds of test items
-Virtualized: More components to be tested
-Diversified services: More patterns to be tested
-Agile implementation: Shorter time for testing
2. Our target and proposal
Policy-based path control-Becoming widely used
(e.g., hybrid WAN)
-Monitor and switch paths
in responding to network
conditions
Testing automation
CloudShell/TestShell-Protocol testing
-Protocol: BGP/OSPF
Does not support testing
in degradation cases
Automatic System for Network Testing with
QualityDegradation
Testing with various quality patterns is needed
3. Automatic system for network testing
NtbT Server
Chef server Zabbix server
, : Virtual Bridge
KVM
OSS DB (testing scenarios,
NtbT topologies, recipes,
logs)
NtbT
(NW to be tested)
Web applicationTesting Control Server
vSW vSWvSW vhost vhost
Physical-SW
-Create test scenarios
-Check results
: Testing Control Plane
: Data Plane
Key Functions for automatic testing
➢ Automatic construction of NtbT
➢ Automatic failure occurrence
➢ Monitoring network status
3.1 Testing scenario
Procedure:
I. Create a testing scenario with start time
II. Scenario runs automatically
III.Check the monitoring result
Testing Scenario No. Scenario Step
NtbT
construction
scenario
1 Selection of NtbT topology from DB
2 Construction of NtbT
3 Initialization of network equipment
4 Checking of connectivity
Failure scenario
5
Selection of recipes from DB with execution time
Traffic recipes : traffic sending/receiving
Failure recipes: interface down/up
latency start/stop
:
IP version
Bandwidth, etc.Node
Interface, etc.
Monitor network
status during testing
3.2 Automatic construction of NtbT
DB
OSS
(Chef, Zabbix)
Testing Control Server
: Chef client
: Chef server
NtbT Server
Construction
Scenario
: Data NW
: Control NW Virtual switch
with Chef client
installed
Create VM
=>Configure
=>Check connectivity
DBJSON-
based
Config.Hostname
Control IF
Data IF,
etc.
3.3 Automatic failure occurrence
DB
OSS
(Chef, Zabbix)
Testing Control Server
NtbT Server
Failure
Scenario
delay
: Chef client
: Chef server
: Data NW
: Control NW
Failure
Recipe Node
IF
Degradation type
Degradation values
Failure
Recipe1SW2
FE0/1
Delay
3000ms
Failure
Recipe2SW2
FE0/1
Delay
4000ms
SW2
FE0/1
Delay
1000ms
Fluctuating
Quality
Failure
Recipe3
3.4 Monitoring network status
snmpd
Testing Control Server vSW, vhost
Traffic volume
Log
DB
iperf
vhost
End-to-End quality(speed, jitter, loss)
zabbix_
sender
zabbix_
server
zabbix_
agentd
Every 10 seconds
Ipef sends/receives/
analyzes packets
every second
↓
Zabbix_sender
uploads periodically
4.1 Network topology for system evaluation
Monitor and Switching
Measured delay(s) State
delay<x NORMAL
x≦delay<y DELAY
y≦delay DOWN
※Threshold x, y is configured on GW.
Monitor & switching are
enabled on GW1-CE1
and GW3-CE1
The path of a state better than another becomes the working path
4.2 Network implementation
OSS
(Chef, Zabbix)
Testing Control Server
NtbT Server
DB
ProxyProxy
Proxy
Proxy
Proxy
Proxy
PC1CE1
L3SW
GW1
GW3
RR GW2 CE2 PC2
OSS client
cannot be
installed
Proxy
for CE2
snmpd/
zabbix_agentd
: Chef client
: Chef server
: Data NW
: Control NWTotal 7 proxy nodes
on the testbed
OSS client
installed for
automation
Testing Control
ServerNtbT Server
OS Ubuntu 14.04
Memory 64GB
CPU 28core 28core
Chef Server 12.4.0 Client 12.5.1
Zabbix 2.4.8 (server) 2.2.2 (agent)
System environment
4.3 Failure scenario for system evaluation
Recipe
No.
Failure Recipes
[x = 3, y = 5]
1 Cause 2 sec delay on the link between GW1-CE1
2 Change the delay with 4 sec
3 Change the delay with 6 sec
4 Normalize the delay
5 Cause 2 sec delay on the link between GW3-CE1
6 Change the delay with 4 sec
7 Change the delay with 6 sec
8 Normalize the delay
GW1-CE1 GW3-CE1Working
path
NORMAL NORMAL A
DELAY NORMAL B
DOWN NORMAL B
NORMAL NORMAL B
NORMAL NORMAL B
NORMAL DELAY A
NORMAL DOWN A
NORMAL NORMAL A
Verify with GW log
Verify with traceroute (manual) or
visualized traffic on system GUI (with system)
Expected Results
4.4 Evaluation results
Perform the scenario manually and, with the proposed
system, twice, respectivelyStep Manual operation With system
1 Execute command Create scenario
2 Check result Run scenario
3 Repeat 1-2 Check result
Case Testing OperationAverage
Time[min]
Average
Work Hours[min]
Case-1 Manual operation 21:05 21:05
Case-2 With system 27:51 11:12
Results
Reduced
46.9%
Operator’s
work hours
5. Conclusion
Proposal • Automatic system for network testing with quality
degradation
Implementation of automatic system• Construction of NtbT and failure occurrence by Chef
• Monitoring network status by Zabbix
System evaluation• Testing on experimental network that consists of IP-Sec GW
• Reduce testing workload by half
• No human error risk
Ref.1) Creating a testing scenario
Execution
timeRecipe Node Parameter
Traffic receiving start
Traffic sending start
Degradation start
Degradation change
Degradation change
Interface
Traffic parameter
(IP version,
bandwidth etc.)
Degradation parameter
(Loss rate,
Latency values etc.)
Ref.2) The scenario running